[MacGallant]

I read an article from a Mac news website about developers with Macintel dev machines claiming that those Pentium 4 3.6 Ghz Macintel beats the pants off their PowerMac Dual G5 2.0 GHz with ease which is very noticable esp. surfing the net with Safari because Safari isn't multi-threaded so the Dual G5 is actually at a disadvantage.

My question is: Is it true that Safari is not multi-threaded? Apple has been singing the praises of dual processing for years , if what those developers said is true, so why isn't Safari multi-threaded? Apple is the posterboy for dual processing.

[Catfish_Man]

It's because a) it's hard, and b) Apple didn't write the original code for the Safari renderer. Safari uses multiple threads (that's easy to check), but it does the rendering on one thread. There's an experimental (aka broken) multithreaded image decoder in Safari 2.

[Thinine]

The difference between Safari on a dual 2 and even the 3.6 GHz Mactel should be negligible, except for the most CPU bound tests. Real world tests are bandwidth limited.

[Catfish_Man]

I disagree. If that were true then Omniweb, Safari, Camino, and IE should all be the same apparent speed. Since I can see clear differences between some of them... bandwidth doesn't appear to be the sole limiting factor.

[Millennium]

Safari is multithreaded. Almost all network apps are, because if they weren't then you wouldn't be able to do anything while you were waiting for network traffic to come in.

The question is, is anything other than the networking aspect threaded? That's hard to say. Some kinds of tasks lend themselves better to threading than others, and so there's not much way to tell without looking at the code. Then again, WebKit is open-source, so you could probably examine the code to find out stuff like that.

[Chuckit]

I'd be surprised if the difference is that noticeable. Surfing the Web doesn't exactly pin the processor on my G4 PowerBook, so it doesn't really make sense for the processor to be that much of a bottleneck on a G5.

[Thinine]

Originally Posted by Catfish_Man

I disagree. If that were true then Omniweb, Safari, Camino, and IE should all be the same apparent speed. Since I can see clear differences between some of them... bandwidth doesn't appear to be the sole limiting factor.

They're different speeds because they all use different rendering engines. The same Safari on a dual 2 and a 3.6 GHz P4 should be indistinguishable for most uses.

[Millennium]

For every article that claims such massive speed gains, there is an article which claims the opposite. Truth be told, we're unlikely to know anything for certain for a long time indeed, because posting any form of objective measurement would be in violation of Apple's NDA.

[Big Mac]

Certainly Safari is multithreaded - it has a bunch of threads open. Just look at the Activity Monitor. The report about Safari being so much faster on Apple's PC developer boxes has to be viewed as ridiculous. If microprocessor performance were the constraining factor, then Safari would be pegging your Mac's processor right now. Safari barely hits my processors when accessing pages. As others have surmised, the most likely cause of the reported speedy performance of these PCs is their high performance drives. Remember those early "developer" reports that claimed G5s were running Virtual PC at native PC speeds? Arm yourself with common sense.

[TETENAL]

Originally Posted by Big Mac

Remember those early "developer" reports that claimed G5s were running Virtual PC at native PC speeds?

No, I don't remember them. VirtualPC didn't run on G5s.

[Big Mac]

Originally Posted by TETENAL

No, I don't remember them. VirtualPC didn't run on G5s.

There were such rumors, and the fact that VPC didn't run at all at that time is the reason why I placed the .